Source-linked AI summary
Close encounters in a pediatric ward: measuring face-to-face proximity and mixing patterns with wearable sensors
L. Isella, M. Romano, A. Barrat, C. Cattuto, V. Colizza, W. Van den Broeck, F. Gesualdo, E. Pandolfi, L. Ravà, C. Rizzo, A. E. Tozzi
TL;DR
Limited information on hospital contact patterns has constrained infection-prevention guidance beyond general and qualitative recommendations. The study measures and analyzes close-contact duration by role and individual in a hospital setting, finding limited patient-patient interaction alongside frequent and long interactions involving nurses and caregivers.
Problem
Limited information on objectively measured hospital contact patterns has constrained preventive-strategy guidelines to general and qualitative recommendations.
Method
The study measures and analyzes the duration of close contacts by different roles and at the individual level in a hospital setting using active Radio-Frequency Identification Devices.
Results
The study found very limited interaction between pairs of patients, while patients had frequent, long contacts with caregivers and nurses showed frequent and long contact patterns.
Takeaways & Limitations
Measuring close-contact duration by role and individual provides information for analyzing contact patterns in hospital settings.
Takeaways & Limitations
Contact collection was not based on objective measurements and was generally performed on a random day.
Abstract
from arXiv · showhide
Nosocomial infections place a substantial burden on health care systems and represent a major issue in current public health, requiring notable efforts for its prevention. Understanding the dynamics of infection transmission in a hospital setting is essential for tailoring interventions and predicting the spread among individuals. Mathematical models need to be informed with accurate data on contacts among individuals. We used wearable active Radio-Frequency Identification Devices to detect face-to-face contacts among individuals with a spatial resolution of about 1.5 meters, and a time resolution of 20 seconds. The study was conducted in a general pediatrics hospital ward, during a one-week period, and included 119 participants. Nearly 16,000 contacts were recorded during the study, with a median of approximately 20 contacts per participants per day. Overall, 25% of the contacts involved a ward assistant, 23% a nurse, 22% a patient, 22% a caregiver, and 8% a physician. The majority of contacts were of brief duration, but long and frequent contacts especially between patients and caregivers were also found. In the setting under study, caregivers do not represent a significant potential for infection spread to a large number of individuals, as their interactions mainly involve the corresponding patient. Nurses would deserve priority in prevention strategies due to their central role in the potential propagation paths of infections. Our study shows the feasibility of accurate and reproducible measures of the pattern of contacts in a hospital setting. The results are particularly useful for the study of the spread of respiratory infections, for monitoring critical patterns, and for setting up tailored prevention strategies. Proximity-sensing technology should be considered as a valuable tool for measuring such patterns and evaluating nosocomial prevention strategies in specific settings.
Introduction
Hospital infection-prevention guidance lacks fine-grained empirical contact data tailored to specific wards. The study proposes wearable RFID sensors to measure individual face-to-face proximity and inform transmission models and prevention strategies.
- Contact patterns are essential for understanding infection transmission and designing targeted interventions.
- Self-reported contact studies rely on interviews and recall, lack objective measurement, and are commonly collected on a random day without longitudinal information.
- Hospital settings require high-resolution spatial and temporal data to characterize interactions objectively and non-obtrusively.
- Limited contact information has kept hospital-acquired-infection prevention guidance general and qualitative, unable to reflect heterogeneity across wards and procedures.
- Fine-grained contact data can inform and validate agent-based nosocomial-infection models and support tailored containment measures.
- The study uses wearable active RFID devices to obtain accurate, tailored individual contact estimates in a pediatric hospital ward.
Methods
The study monitored face-to-face proximity among ward participants using wearable RFID badges and a distributed reader network. Contacts were defined and aggregated at 20-second intervals for category- and individual-level analysis.
- The study took place in a 44-bed general pediatric ward arranged as 22 two-bed rooms.
- Participants included ward assistants, physicians, nurses, patients, caregivers, tutors, accompanying persons, and visitors.
- Wearable active RFID devices exchanged low-power packets, while ward readers, a LAN, and a central computer collected and stored proximity data.
- The deployment was configured to detect face-to-face proximity within 1–1.5 meters when badges were worn on participants’ chests.
- A contact was recorded in a 20s interval when paired devices exchanged at least one packet at the lowest power level, continuing until an interval without exchange.
- The analysis measured contact occurrence, frequency, encounter duration, cumulative contact time, and corresponding individual-level aggregates.
Results
The pediatric-ward measurements revealed role-specific mixing patterns, with dense healthcare-worker interactions but highly specific patient-caregiver contacts. Individual-level networks reinforced the limited mixing among patients and caregivers.
- 119 individuals were included in the analysis after participant and signal-quality exclusions.
- Approximately 20 contacts per participant per day were recorded, with ward assistants and nurses involved in especially large fractions of contacts.
- Patients and caregivers each represented approximately 22% of total contacts, while physicians represented 8%.
- Patient-caregiver encounters were frequent but involved few distinct partners, consistent with strong one-to-one interactions.
- Long interactions occurred between patients and caregivers, among nurses, and among ward assistants, whereas patient-patient and visitor-visitor proximity time was extremely small.
- Healthcare-worker networks were dense, while each patient and caregiver interacted essentially with one corresponding partner and had few contacts within their own classes.
Discussion
The study provides direct contact measurements that capture heterogeneous, class-specific interaction patterns in a pediatric hospital, informing infection-transmission models beyond homogeneous assumptions. Patient-caregiver contacts were highly specific, whereas nurses occupied a central position in frequent and prolonged interactions, supporting targeted prevention strategies.
- Contribution: Direct contact measurements capture heterogeneity in hospital interactions and can parameterize infection-spread models beyond simple homogeneous assumptions.RFID-based measurements account for heterogeneities among patients, health care workers, and visitors and provide inputs for mathematical and computational models.
- Interaction patterns: Patients and caregivers had very limited interactions with one another, while physicians had few, short contacts with patients.The overall interaction structure showed strong specificity in patient-caregiver contacts contrasted with health care worker interactions.
- Interaction patterns: Patient-caregiver interactions were intense and continuative but mostly involved the corresponding patient, limiting caregivers’ potential to spread infection widely across the ward.The findings support focusing prevention resources on caregiver-patient interactions rather than all possible caregiver contacts.
- Interaction patterns: Nurses had frequent and prolonged contacts with patients and with one another, supporting their prioritization in local infection-control interventions.The authors associate this intensive professional contact pattern with potentially higher airborne-infection risk among nurses in the studied setting.
- Limitations: The one-week deployment and strict infection-control conditions limit assessment of whether the observed patterns remain stable over time.The authors call for longer and more extensive deployments to characterize expected behaviors and fluctuations and assess temporal stability.
Figures
The figures define how face-to-face contacts are represented and summarize contact frequency, duration, role-pair mixing, and cumulative interaction networks. Supplementary matrices assess whether these patterns remain robust without RFID filtering.
- Figure 1: Figure 1 represents individuals as nodes and detected contacts as links, including repeated links for recurrent interactions.For the example pair N1–P1, one contact occurs twice and lasts six minutes; N1 has three contacts involving two distinct contacts and seven minutes total.
- Figure 2: Figure 2 compares median contacts, distinct contacts, and cumulative contact time per participant across roles, normalized to 24 hours.Each role’s quantities include contacts with any other participant.
- Figure 3: Figure 3 shows role-specific probability distributions for individual contact counts and cumulative contact time, with duration normalized to a 24-hour interval.The distributions include contacts established by each individual with any other individual.
- Figure 4: Figure 4 displays non-zero cumulative contact-duration distributions for role pairs using medians, interquartile boxes, 5th–95th percentile whiskers, and outliers.After normalization, the 20-second experimental resolution appears as a lowest visible value of 2.5 seconds.
- Figure 5: Figure 5 encodes median role-pair node strengths in grayscale for contact counts, distinct contacts, and cumulative contact time.Rows and columns represent role classes; zero-strength individuals contribute to the corresponding medians, and matrix asymmetry reflects unequal class sizes.
- Figure 6: Figure 6 presents cumulative contact networks for pairs of classes and interactions within each class, with nodes as individuals and edges weighted by cumulative face-to-face time.Off-diagonal nodes are ordered by increasing number of edges.
- Supplementary Figure S1: Figure S1 repeats the contact matrices without RFID-badge filtering and reports robustness relative to Figure 5.The same three metrics are shown with grayscale-coded median values and 24-hour-normalized contact durations.
Tables
The tables document the analyzed sample and summarize participant classes and contact measurements. Table 1 contrasts included and excluded patients, while Table 2 organizes contact totals and participant-level daily contacts.
- Table 2: Table 2 summarizes study classes, contact counts, total daily contacts by class, and median daily contacts per participant.The total daily contact measure is the sum of contacts established by individuals in each class.